Week 5 Summaries

Going out: Robust Model-based Tracking for Outdoor Augmented Reality:
The paper presents an augmented reality system that provides realtime overlays on a handheld device.  Traditional augmented reality systems rely on GPS for outdoor position measurements and magnetic compasses and inertial sensors for orientation.  In urban outdoor environments GPS is hindered by buildings and signal reflections.  The system described in this paper uses a computer vision based solution to help with where GPS falls short.  Outdoor tracking systems have bigger obstacles to overcome due to carrying lighting and bad viewing conditions.  The work in this paper combines edge-based trackers with inertial sensors to help eliminate errors with fast moving objects.  The system was tested by implementing a location based game using a handheld augmented reality platform.  Initialization was one problem that was left for future work.  In the experiment the user playing the game must start in a well defined known location because of the inaccuracies of GPS.  The tested system operated at about 15-17 frames per second where OpenGL incurred the largest overhead, so newer graphics hardware would speed up the system.

KinectFusion: Real-time 3D Reconstruction and Interaction Using a Moving Depth Camera
The paper describes KinectFusion a technology that can rapidly create 3D models of indoor objects and spaces.  The KinectFusion uses the Kinect camera to generate real-time depth maps and can construct a scene in seconds.  The biggest benefit of KinectFusion is the low-cost of the scanner.  After a room is scanned software can be used to introduce virtual objects with physics onto the scanned real-world objects.  KinectFusion brings new possibilities with what can be done in the space of object-scanning, augmented reality, and physics based interactions, all at a low price point.

Comments are closed.